Search Visibility: New Ways to Rank in Google, AI Search, and LLMs
Hitting the top today takes more than old-school SEO. Keywords, backlinks, and on-page SEO still matter, but they’re not enough. Google keeps its ranking mechanics under wraps, while AI is shaking up the market. AI answers are already built into Google Search, ChatGPT alone draws 900 million weekly active users, and over a third of people now kick off their searches with AI.
At PropellerAds, we know traffic inside out, helping businesses reach massive audiences of real users across global markets. Organic visibility runs by different rules, so we handed the mic to Oleg Shestakov, below is his account of the methods circulating in the SEO world today.
What follows is Oleg Shestakov’s account of what the SEO scene is testing right now, a field report, not a playbook. Anyone experimenting does so at their own risk.
How Google visibility works
And we start with Google. Google usually keeps its cards close. Cut through the noise, though, and here’s what practitioners believe matters.
According to publicly reported Content Warehouse API documentation, these factor names have been discussed in the trade press since 2024. Google has not confirmed that the documents describe live ranking behavior, and their current relevance is a matter of industry interpretation rather than established fact.
| Trust Google evaluates overall site authority. | CTR Google treats a lack of clicks as a negative signal | Backlinks Google evaluates relevance, the authority of the linking page, and anchor text. |
| Match Search Intent Google evaluates not just clicks, but successful clicks | Long Clicks Google considers the expected length of user engagement. | Host Age New sites face ranking limits |
- Trust/Authoritative site. The higher a site’s authority, the more tolerance it tends to have in rankings. Google collects a range of signals to assess trustworthiness. These are content quality, relevance, security, site reputation, and trusted users with real browser history.
- CTR. Optimize your title and description. Nail the query match; keep the promise in the snippet consistent with what the page actually delivers.
- Backlinks. Quality beats quantity. Go after backlinks from strong, relevant pages. Links from reputable media are considered trustworthy by default.
- Match Search Intent. Users should get exactly what they wanted to find. Remove slow-loading, intrusive pop-ups. Keep users engaged with clear navigation and related content.
- Long Clicks. Search → click → stay → read → interact. That’s a good user journey.
- Host Age. New sites face limited visibility for some time – this is a Google anti-spam mechanic. Shestakov’s advice here is to get a site indexed early and let the domain age before a project launches.
Tactics for Google
Now for the interesting stuff. Oleg Shestakov shares how some practitioners ride a sudden wave of buzz to give their sites a boost in Google’s rankings.
Turn a traffic spike into profit
The description in this section is reported for information only. Buying traffic in order to simulate organic demand and influence search rankings is contrary to Google Search Essentials. PropellerAds does not offer its inventory for this purpose and does not recommend this approach.
Google loves a good hype spike, really. How else do you think a random celeb video blows up worldwide overnight? A sudden surge in popularity isn’t bad itself – as long as it comes from genuine user interest. That condition is the whole point: a bought impression and a visit that starts with the user’s own intent are different events, and only the second is what this mechanic rewards.
So, users must be real. We mean, REAL. Browser-level signals, session history, and user behavior can help distinguish genuine visitors from automated traffic, so experts focus on real audiences.
What Else Can Get a Boost?
It’s not just for standard organic pages. This approach stretches further:
- images;
- YouTube and TikTok videos;
- third-party hosted pages.
How Far Can a Popularity Spike Push a Site?
Sites already in the top 30 are the easiest to move. These are the figures Shestakov reported from his own testing; the sample, period, and methodology were not disclosed to us, and we have not verified them:
- Top 30 → Top 10: happens almost daily;
- Top 150 → Top 10: also quite possible.
In his experiments, the typical lift lasted around 24 hours before the rankings started to slip back. In 10-15% of cases, however, the new position stuck for much longer – weeks, months, or sometimes even permanently. There was no way to predict when that would happen.
How LLM visibility works
LLM traffic can convert like crazy, but most people have no idea what drives it. Together with Oleg, we’re breaking it down.
LLMs pull their sources from different search ecosystems. Shestakov’s own estimate of the split, which we have not been able to verify against a published source, is:
- 60% from Google;
- 40% from Bing.
LLMs are probabilistic, so results change. Sometimes ChatGPT shows a source that doesn’t even exist in Google’s index. There’s no one-size-fits-all formula for getting picked, but AI models do share a few basics that make content easier to extract.
Six things that decide whether a model pulls your page into an answer, and the types of sources where each one matters most.
- Authoritative portals
- Niche sites and personal blogs
- Reddit & Quora
- Comparison and ranking pages
Relevance to the query from the very first paragraph. The model should immediately understand: “This page has the answer I need.”
No model digs deep into the text, so the first words carry the most weight. Don’t bury the answer under a long intro.
One URL equals one clear intent. Keep it apples to apples and don’t mix unrelated topics. If the intent is fuzzy, the model will skip the page.
Add a clear Q&A or summary block so the model can easily find a question with a ready-to-use answer next to it.
Experience, expertise, authoritativeness, and trustworthiness.
Content should flow from simple to complex so the model can extract information step by step.
What each LLM wants
We’ve just mentioned that different AI models see and evaluate information differently. So what does that actually mean? And yes, it’s easy to get lost in the flood of AI models out there. Relax – most are noise.
These are the ones worth focusing on, according to Oleg:
| Criteria | ChatGPT | AI Overviews Google Search AI answers | Google AI Mode | Perplexity |
|---|---|---|---|---|
| Early query confirmation | ||||
| Intent Match | ||||
| Early Query Answer | ||||
| Sections Coherence | ||||
| Summary | ||||
| E-E-A-T | ||||
| Author Credibility | ||||
| From Simple to Complex |
The ideal source for LLMs: what’s inside?
Extractability is everything. The goal isn’t to impress the reader. It’s to make it easy for the model to quickly pull out the facts.
Bigger tag, bigger effect on whether your page gets quoted.
More content = more chances to get picked. Shestakov’s view is that a larger number of smaller publications can outperform a couple of placements on a major portal – a different strategy from the authoritative-portal route described below, not a replacement for it. Note that mass-producing near-identical pages across many sites is treated as scaled content abuse.
Tactics for LLM Visibility
Now, let’s look at a few techniques modern SEO practitioners use to get content directly into LLM results.
Why LLMs pick up rankings
Getting into AI through authoritative portals is the easiest way. Those with the budget can simply buy articles and ads there – at a pretty insane price. For example, to get published in Forbes, you should either qualify for Forbes Councils membership or take the BrandVoice route. Paid placement programs on major business media are generally reported to run into the tens of thousands of dollars [source, date].
Either way, getting in is neither cheap nor easy. Niche sites – sit and hope that one day an LLM turns to you. Taking part in the relevant Reddit and Quora threads is another route – but only with the commercial affiliation disclosed openly.
Undisclosed promotion and vote manipulation breach those platforms’ own rules. But rankings? They hit hard – for real. Models evaluate structure and usefulness. A clean, well-organized ranking is easy to parse, so they pick it as a source.
One approach used in modern SEO is to build rankings specifically with LLM retrieval in mind:
- Ranking criteria. Top-ranking sites can reveal which criteria and comparison points are commonly used.
- A query fan‑out. A broad query is broken down by an LLM into narrower, related searches.
For example, “best movie reviews” might produce: “best movie review sites”, “most trusted movie review sites”, “best movie critics”, and “best websites for movie ratings”.
- Those queries become H2s. Each one gets a short, focused paragraph.
- Brands are ranked and compared. A brand may take the top spot, with competitors below it. The comparison must rest on objective, verifiable criteria, must not disparage the competitors named, and the page must state plainly who publishes it.
- An FAQ adds direct answers and useful brand facts. AI loves this format because it’s easy to extract.
- Ranking should then be connected to genuinely independent third-party sources that already report the same facts – not to placements the ranking’s own publisher controls.
- The ranking goes live.
- The approach can be automated and scaled. More sources = more chances to get picked. Scaling it also scales the disclosure obligation above, and automated mass production of near-identical ranking pages is treated as scaled content abuse.
Increasing LLM visibility directly
Here’s another trick that has the SEO community buzzing. The idea is to bring real activity around a brand into the AI ecosystem and reach users who rely on AI for search. Driving purchased traffic into a third-party AI service in order to influence what that service retrieves is contrary to the terms of use of the services concerned.
Like this: Prompt → AI search/retrieval → Your site. Sounds good, doesn’t it?
Take a fictional site called BessstRankings. The approach is built around a simple algorithm:
- A clean session. An incognito window minimizes personalization.
- Prompt testing. Relevant audience questions are tested to see when the AI mentions the site or retrieves related information.
- A query URL. A relevant prompt is embedded into an AI service URL.
- A banner campaign. The query URL can then be used as the banner destination, letting interested users find the site through AI search.
Here are questions real listeners asked our guest speaker.
F.A.Q.
It matters who imposed the ban. If it’s Google, you need to fix the issues behind it first. If the domain still doesn’t recover once the underlying violations have actually been removed, rebuilding on a fresh domain with updated, compliant content may be an option, with a 301 redirect used for a proper migration. Moving to a new domain without fixing the original problems is circumvention of the platform’s enforcement, not a migration.
Some of them already are limited. Big platforms also keep their algorithms under lock and key, so anything that works today can stop working without notice. The risk sits entirely with whoever applies these tactics.
No guarantees; it’s random. In theory, it is worth testing when the sources in your niche are stale. If the pages being picked are from 2024-2025 and nobody is keeping them fresh, that could be the decision. One caveat that is not optional: paid placements and paid links must be marked rel=” sponsored” and disclosed as paid.
This post reports what Oleg Shestakov shared on stage. We do not endorse the tactics described; anyone applying them does so at their own risk.
Join our Telegram for more insights and share your ideas with fellow affiliates!
